About & FAQ
Everything you want to know about AI employees.
I built StartupFormulas because the AI install work wasn’t being done. Operators were landing acquisitions, the agents were real, and what was missing was a partner who arrived after the wire cleared, picked the right tools, and stayed long enough to train the team. The questions below cover everything I get asked before someone decides whether to work with us.
— Nathan, founder
Reach me at startupformulas.com/get-started
AI Employees
- What is an AI employee?
- An AI employee is a software agent that performs a specific job role autonomously — answering inbound calls, chasing unpaid invoices, following up stalled proposals, or monitoring subscription churn. Unlike a general chatbot or AI assistant, an AI employee has a defined scope, an accountability metric, and is installed into your specific business systems. StartupFormulas installs fifteen named AI employee roles for SMB operators in trades, services, and property management.
- How are AI employees different from AI chatbots?
- A chatbot responds to questions. An AI employee initiates work, runs a scheduled cadence, and produces a measurable output — a recovered invoice, a booked appointment, a collected tax document. The distinction is accountability: an AI employee owns a job, not a conversation. It works whether you're in the office or not and reports back on what it did.
- Do AI employees replace human employees?
- Mostly no, sometimes yes — and we'll tell you which one you are up front. AI employees absorb the work you currently can't hire for: the 6pm callback, the 73-day AR chase, the proposal follow-up nobody owns. If you've been carrying an open requisition for six months, an AI employee replaces that hire and the math finally pencils. If you've got a strong office manager already doing the work, the AI employee gives her her week back so she can focus on what only humans do — relationships, judgment calls, the hard customer save. We name which scenario you're in during the Week 1 audit. Honesty about replacement is part of why operators trust the install.
- What types of AI employees does StartupFormulas install?
- StartupFormulas installs fifteen named roles across four business levers. Acquisition: AI Receptionist, AI Marketing Manager, AI Outbound SDR, AI Sales Estimator. Retention: AI Customer Success Manager, AI Tax-Season Coordinator. Recapture: AI Collections Specialist, AI Reactivation Specialist, AI Proposal Follow-up Rep, AI Service-Reminder Coordinator. Pricing: AI Competitive Pricing Optimizer. Each role has a named outcome metric, a human-salary anchor, and a typical install timeline.
- How long does it take to install an AI employee?
- The median install time is 27 days from kick-off to live. Simple installs — AI Receptionist, basic AR dunning — typically go live in 14 to 21 days. More complex roles requiring deep CRM integration or compliance review take 45 to 60 days. Every install starts with a stack audit in the first week so there are no timeline surprises.
- What results can I expect from an AI employee?
- Results depend on the role and your starting numbers. Representative outcomes: an AI Receptionist cuts voicemail rate from 30% to under 5% within 30 days; an AI Collections Specialist pulls $80K of stuck AR current within 90 days; an AI Customer Success Manager reduces annual churn from 24% to 18% within 90 days; an AI Tax-Season Coordinator frees 80+ partner hours per season. These are illustrative ranges drawn from typical operator data, not guarantees.
- What industries do AI employees work for?
- AI employees work best in HVAC and mechanical contracting, roofing and exterior trades, property management, bookkeeping and accounting firms, marketing agencies, and B2B service businesses. The best-fit revenue band for a managed install is $5M to $15M — multi-location operators and scaling owner-led businesses where the hiring pain is sharpest. We also serve $2M to $5M owner-operators on a smaller install, and post-acquisition holdcos as a separate engagement track. The common thread is high-repetition, rules-based work with measurable volume: calls answered, invoices collected, proposals followed up.
- What businesses are not a good fit for AI employees?
- VC-backed tech startups with engineering teams should build their own agents. Businesses under $2M revenue typically don't have enough throughput to justify a $15K+ install — start with the $99 skill instead, which tells you exactly when an install would pencil. Businesses in heavily regulated verticals — medical billing, legal practice, certain insurance lines — require compliance review before any install. If you want a strategy deck rather than a working system, this is not the right fit.
AI Agents & How They Work
- What is an AI agent?
- An AI agent is software that can perceive its environment, make decisions, and take autonomous actions to achieve a goal — browsing the web, reading email, calling APIs, executing code. Unlike a large language model responding to a single prompt, an agent runs multi-step workflows and can operate without a human approving every step. AI agents are the underlying technology that powers AI employees.
- What is the difference between an AI agent, an AI assistant, and an AI chatbot?
- An AI chatbot responds to questions inside a conversation interface. An AI assistant responds to user requests and may take simple actions like setting reminders or summarizing documents. An AI agent executes multi-step autonomous workflows — reading your AR aging report every Friday, deciding which invoices to pursue, sending the dunning emails, logging the results — without a human triggering each step. An AI employee is an AI agent with a specific job role, accountability metric, and defined scope.
- What is an AI agent skill?
- An AI agent skill is a structured instruction set — typically a markdown file — that tells an AI agent how to behave in a specific domain. Think of it as a job description the AI reads at runtime. The Nathan Method is a skill: when loaded into Claude Code, Cursor, Codex, or Gemini CLI, it gives the agent the scoring rubric, intake protocol, and install playbook for SMB business operations. It follows the open Agent Skills standard and works with any compatible agent platform.
- What AI platforms do StartupFormulas AI employees run on?
- Depending on the role, AI employees are built on a combination of: voice AI platforms — Bland.ai, VAPI, Retell — for receptionist and phone-based roles; workflow automation — Make, n8n, Zapier — for scheduling, dunning, and document collection; outreach platforms — Clay, Apollo, Instantly — for sales roles; and LLM providers — Anthropic Claude, OpenAI GPT, Google Gemini — for reasoning and writing tasks. The right vendor stack is selected during the stack audit, specific to your business.
- What is the Nathan Method and how does it use AI agents?
- The Nathan Method is a $99 AI agent skill loaded into your AI agent (Claude Code, Cursor, Codex, Gemini CLI, GitHub Copilot, Amp, or any Agent Skills-compatible platform). It runs an 11-question intake, scores four operational levers — Acquisition, Retention, Pricing, Recapture — against your numbers, and names the single AI employee to install first. The output is a prioritized recommendation, a 90-day install calendar, and a weekly KPI. The skill identifies the right hire; the managed install does the actual installation.
- What is the open Agent Skills standard?
- The Agent Skills standard is an open format for packaging AI agent behaviors as portable, composable skill files. A skill is a ZIP file containing a SKILL.md instruction set and optional configuration. Any AI agent that implements the standard can load and execute the skill. The Nathan Method uses this standard, making it compatible with Claude Code, Cursor, GitHub Copilot, OpenAI Codex, Gemini CLI, Amp, and 25+ other agents without modification.
- Is AI ready for SMB business operations?
- For well-defined, high-repetition jobs — answering calls, chasing invoices, following up proposals, collecting documents — AI is production-ready today. For nuanced client relationships, complex negotiations, and high-judgment work, humans remain essential. Installs fail when scope is too broad. Installs succeed when they own one specific job with a single measurable output. We narrow scope before installing anything.
The Install & Process
- How does the Nathan Method skill work step by step?
- You purchase the $99 skill, unzip it into your AI agent's skills folder, and load the skill. Type /intake to run the full 11-question intake — or /goal followed by your desired outcome to skip the intake. The skill scores the four levers, names the AI employee to install first, generates a 90-day install calendar, and gives you one KPI to watch every Friday. The process takes 18 minutes in a single session.
- What is the difference between the $99 skill and the full managed install?
- The $99 skill is the DIY path: you install the skill into your agent, run the audit, and get the roadmap and recommendation. You still execute the AI employee install on your own systems. The full managed install ($15K to $35K) is the done-for-you path: we conduct the stack audit, select vendors, integrate the AI employee into your infrastructure, train your team, and run the 90-day pilot. Same underlying Nathan Method. Different price. Different hands on the wheel.
- What does a managed install include?
- A managed install includes a Week 1 stack audit and team interviews, a vendor selection memo naming the specific tools and contracts, AI employee integration on your existing infrastructure, operator and team training on working alongside the AI employee, 90 days of managed operation and performance monitoring, and a hand-off playbook your team can run independently after we leave.
- Does StartupFormulas provide ongoing management?
- The standard managed install includes 90 days of active operation and a hand-off. Most clients are operationally independent by day 90. Optional ongoing optimization retainers are available for businesses that want quarterly reviews, expanded installs, or continuous performance monitoring.
- Can AI employees integrate with HubSpot, QuickBooks, or my existing CRM?
- Yes. Most AI employees connect to standard SMB software: QuickBooks and Xero for accounting data, HubSpot, Jobber, ServiceTitan, and GoHighLevel for CRM, Twilio and RingCentral for telephony, and email platforms for outreach. The stack audit in Week 1 maps every integration point before any build begins. If your software is not compatible, we tell you in the audit, not after the install.
Pricing & Commitment
- How much does a managed AI employee install cost?
- Managed installs are priced at three tiers based on revenue band and scope. The Owner-operator tier ($15,000) suits businesses at $2M to $5M revenue with two AI employees installed over 30 days. The Multi-location tier ($25,000) is the best-fit band — businesses at $5M to $15M with four AI employees over 60 days. The Mature operator tier ($35,000) suits businesses at $15M+ with five-plus AI employees over 90 days. All tiers are fixed-fee — no retainer, no hourly billing.
- Is there a retainer or ongoing monthly fee?
- No. The managed install is a one-time fixed fee. After the 90-day pilot and hand-off, you run the AI employees yourself. Optional quarterly optimization retainers are available but not required for the AI employees to keep working.
- Is the $99 skill refundable?
- Yes. The Nathan Method skill has a 14-day, no-questions-asked refund policy. Request the refund through the form at startupformulas.com/get-started with your order ID and we process the refund the same day.
- What if the full install does not produce results?
- We do not guarantee specific outcomes — businesses vary too much for that to be honest. What we guarantee is transparency: if the install is not producing measurable results by day 30 of the pilot, we say so and rearchitect before you pay the back half. The install is only complete when the first output metric has been recorded.
Trust, Results & Who We Are
- Has the Nathan Method been tested on real businesses?
- The Nathan Method is built from patterns across real SMB operations. The use cases on this site are clearly labeled as illustrative — they describe how installs run in each vertical using composite data. Named, signed-off case studies are published as they are approved. We publish synthetic use cases so you can evaluate the methodology before buying rather than taking it on faith.
- How is StartupFormulas different from an AI consultant or agency?
- Consultants give you a strategy deck. Agencies manage your existing campaigns. We install working AI employees that produce measurable output from week one. The engagement ends when your team can run the AI employees without us — not when the retainer ends. We charge for the install, not the hours.
- How is StartupFormulas different from buying an AI tool directly?
- Vapi, Lindy, and Sintra sell you the raw AI. You still have to pick which role to install first, configure the workflow, wire it into your CRM, and train your office manager. We do those four steps for a fixed fee in 30 to 90 days and leave you with a working install plus the playbook. Tool vendors sell software. We install employees.
- Should I get an enterprise AI platform like Workday's Agent System of Record instead?
- Only if you already have multiple AI agents running and need to govern them at scale. Workday's Agent System of Record (and similar enterprise platforms like Salesforce Agentforce or Microsoft Copilot Studio) is built for enterprises consolidating fragmented agent deployments — what they call 'agent sprawl.' If your business has zero AI agents today and you're trying to figure out which one to install first, an enterprise platform sells you the meta-layer before you've made the hire. We do the opposite: name the role, install it, hand off in 90 days. When you eventually need governance at scale, you'll know — and you'll have install experience to evaluate enterprise platforms with real context.
- Is my business data secure when using AI employees?
- AI employees only access the systems and data you explicitly authorize during the install. Every data access point is documented in the vendor selection memo before any integration begins. AI employees do not train on your data — they use your data to do their job but do not feed it back into model training. For full details, see the Security page.
- Who runs StartupFormulas?
- StartupFormulas is founder-led by Nathan. The firm exists because the AI install work was not being done — acquirers were landing, the agents were real, and what was missing was a partner who arrived after the wire cleared, picked the right tools, and stayed long enough to train the operator. The methodology, the skill, and the installs are Nathan's direct work.
How we work
Six things that are true about every install.
01
We do the work the agencies refuse.
The McKinsey decks are written. The frameworks are named. The boutique AI consultancies have already pitched you. We're the ones who arrive on day 91 and integrate the actual tools.
02
We don't sell software.
We pick the right vendor stack for your specific business. If Jobber is right, we install Jobber. If GoHighLevel is right, we install GHL. The honest answer is almost never 'our product.'
03
We finish.
Most agencies leave the operator half-trained on a system they don't fully understand. We're done when your team can run it without us — not when the retainer ends.
04
We charge for the install, not the hours.
Fixed-fee install. Managed monthly optimization available when the workflow needs ongoing monitoring, tuning, reporting, or campaign management — your call, not a default. The contract is the deliverable, not the timesheet.
05
We don't believe the agents are smart.
We believe the install is repeatable. The reason this works isn't AI — it's that someone is finally accountable for the integration.
06
We tell you the math.
Every recommendation comes with the cost, the integration time, the operator hours, and the expected lift. You see the same spreadsheet we do.
Still have questions?
The scope call is free. Twenty minutes, honest go/no-go, and I’ll answer anything the FAQ didn’t cover.